Am 31.10.2016 um 17:15 schrieb Bald Eagle:
> Well *** GOLL-LY *** !!!
> Thanks for that 'little' look under the hood  :O
> 
> It's always interesting to see things from a different perspective and be
> reminded that what one sees in a high-level language like SDL has no relation to
> what takes place on the lower levels.

As a matter of fact it's also interesting for me, because I rarely
examine such matters in such excessive detail. (My perfectionism tends
to reach its peak when communicating with other people.)

For example, as I had already hinted at, I had never before paid much
attention to the per-thread temporary data.

I'm sure there must be a way to significantly trim down that temporary
data store, at least for normal cases.

>> the total cost can be
>> minimized by splitting up your N blob elements among C*sqrt(N) blob
>> primitives (with C being a constant depending on the number of threads),
>> in which case the per-thread cost becomes negligible for large N.
> 
> This would be a nice bit of advanced user info to add to the documentation, and
> perhaps even develop into a macro or something in the source / parser / message
> stream.   Something along the lines of "You have >n spheres, consider .... blobs
> ..... to reduce memory usage...."

I don't think such a warning message would be a good idea; while using
blobs as a bulk sphere surrogate does reduce the memory footprint (if
done right), it certainly will be at the cost of significantly reduced
render speed.

Adding a corresponding section to the docs is another matter.

Post a reply to this message